Roadmap: The Path to 10M+ Rows

[wangjiawen2013]

Charton aims to be a lightweight, high-performance, and versatile Rust visualization library built on the Grammar of Graphics.

1. Decoupling & Core Optimization

• Remove Hardcoded Polars Logic: Refactor internal compute kernels to use standard library primitives, ahash for fast grouping, and rayon for data-parallel execution.
• Conditional Feature Gating: Move Polars support to an optional cfg feature. Ensure the core library remains lightweight and dependency-lean.
• Universal Ingestion: Implement a Dataset struct to unify inputs from Vec, ndarrays, and DataFrames into a normalized internal format.

2. API & Macro Ergonomics

• Implement a polymorphic chart! macro that leverages pattern dispatching to ingest either unified data sources (like DataFrames) or variadic raw vectors with automatic stringify!-based header capture.

3. High-Performance Data Pipeline

• Hybrid Precision Engine: Use &[f64] for internal analytical computations (aggregation, scaling) to maintain precision, and &[f32], bytemuck for GPU-bound rendering buffers.
• Zero-Copy Slicing: Optimize the data flow to ensure that 10M+ rows are processed via non-allocating slices wherever possible.
• png backend: Replace resvg with tiny-skia to implement true png backend.

4. WGPU Backend:
Finalize the pipeline for direct hand-off to GPU vertex buffers, enabling 60 FPS interaction on massive datasets

• wgpu + tiny_skia hybrid architecture
• pure wgpu
• remove lyon for draw_path to obtain the highest performance. Use WGSL Shader to calculate the line thickness instead of lyon.

5. Advanced Analytical Features

• High-Arity Faceting: Implement a robust "Small Multiples" (Faceting) engine based on fast grouping logic.
• Intelligent Decimation: Add optional data downsampling/aggregation strategies for rendering optimization when row counts exceed screen resolution.